Extracellular vesicles and their miRNA cargo in retinal health and degeneration: mediators of homeostasis, and vehicles for targeted gene therapy

Purpose Photoreceptor cell death and inflammation are known to occur progressively in retinal degenerative diseases, however the molecular mechanisms underlying these biological processes are largely unknown. Extracellular vesicles (EV) are essential mediators of cell-to-cell communication with emerging roles in the modulation of immune responses. EVs including exosomes encapsulate and transfer nucleic acids, including microRNA (miRNA), to recipient cells which in disease may result in dysfunctional immune responses and a loss of homeostatic regulation. In this work we investigated the role of isolated retinal small-medium sized EV (s-mEV) which includes exosomes in both the healthy and degenerating retina. Methods Isolated s-mEV from normal retinas were characterized using dynamic light scattering, transmission electron microscopy and western blotting, and quantified across 5 days of photo-oxidative damage-induced degeneration using nanotracking analysis. Small RNAseq was used to characterize the miRNA cargo of retinal s-mEV isolated from healthy and damaged retinas. Finally, the effect of exosome inhibition on cell-to-cell miRNA transfer and immune modulation was conducted using systemic daily administration of exosome inhibitor GW4869 and in situ hybridization of s-mEV-abundant miRNA, miR-124-3p. Electroretinography and immunohistochemistry was performed to assess functional and morphological changes to the retina as a result of GW4869-induced exosome depletion. Results Results demonstrated an inverse correlation between s-mEV secretion and photoreceptor survivability, with a decrease in s-mEV numbers following degeneration. Small RNAseq revealed that s-mEVs contained uniquely enriched miRNAs in comparison to in whole retinal tissue however, there was no differential change in the s-mEV miRNAnome following photo-oxidative damage. Exosome inhibition via the use of GW4869 was also found to exacerbate retinal degeneration, with reduced retinal function and increased levels of inflammation and cell death demonstrated following photo-oxidative damage in exosome-inhibited mice. Further, GW4869-treated mice displayed impaired translocation of photoreceptor-derived miR-124-3p to the inner retina during damage. Conclusions Taken together, we propose that retinal s-mEV and their miRNA cargo play an essential role in maintaining retinal homeostasis through immune-modulation, and have the potential to be used in targeted gene therapy for retinal degenerative diseases.

pathogenic roles in these diseases by promoting angiogenesis 427,428 , and modulating immune responses, including in the recruitment of immune cells 410,418,[429][430][431][432] ; features that contribute to cell death. To date however, the identification and role of exosomes and their miRNA cargo in the retina in both healthy and diseased states, is largely unexplored 222,413 . While small extracellular vesicle fractions isolated following high speed >100,000xg ultracentrifugation and expressing tetraspanin markers CD9, CD63 and CD81 (such as those isolated in this work) are commonly referred to as exosomes, without evidence of endosomal origin and in complying with MISEV 2018 guidelines 238 , are herein referred to as small-to-medium EV, or s-mEV. In reference to other works, EV terminology will be referred to as published in the original papers.
Characterizing the role of s-mEV and their miRNA cargo in both the normal and degenerating retina will aid in elucidating novel cell-to-cell communication pathways that could play a role in propagating inflammation during retinal degenerative diseases.
Furthermore, uncovering the miRNA signature within retinal s-mEV as well as their potential binding partners may reveal novel regulatory mechanisms underpinning retinal degenerations, ultimately leading to the discovery of therapeutic targets. This study characterizes for the first time, retinal-derived s-mEV from both the healthy and degenerating mouse retina using a previously established model of photo-oxidative damage-induced retinal degeneration 265 .
Photo-oxidative damage models such as the one employed in this study accurately replicate key pathological changes seen in AMD, including the upregulation of oxidative stress and inflammatory pathways, progressive centralized focal photoreceptor cell loss and microglial/macrophage recruitment and activation 264,265,272,273 .
We show that s-mEV secretion is inversely correlated to photoreceptor survivability, with the severity of retinal degeneration directly correlating to decreased retinal s-mEV numbers. We used small RNAseq to characterize the miRNA cargo of retinal s-mEV (exoMiR). Although we demonstrated that there was no change in s-mEV miRNA-cargo in response to retinal degeneration, we found that s-mEVs contain a set of uniquely enriched miRNAs. Further, we show that miRNA contained in retinal s-mEV were associated with the regulation of inflammatory, cell survival and motility pathways. Upon systemic exosome inhibition using GW4869, retinal function in healthy and photo-oxidative damaged mice was significantly reduced compared to controls. In addition, photoreceptor cell death and inflammation were significantly increased in GW4869-injected photo-oxidative damaged mice, compared to controls. Using in situ hybridization, we further demonstrated that the expression of miR-124-3p in the inner retina was reduced in GW4869-injected photo-oxidative damaged mice, suggesting that miR-124-3p movement could be mediated through s-mEVdependent transport.
Taken together, these results suggest a novel role for s-mEV and s-mEV-miRNAmediated cell-to-cell communication in the retina. We demonstrate that maintaining and transporting necessary levels of s-mEV cargo is required for normal retinal homeostasis and immunomodulation, with insufficient bioavailability of s-mEV potentially leading to inflammatory cell death (see Figure Error! No text of specified style in document..10). In addition, this work elucidates downstream biological targets of s-mEV-miRNA that are required for retinal homeostatic maintenance, and identifies potential miRNA and mRNA therapeutic targets for further investigations.

Animal handling and photo-oxidative damage
All experiments were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and with approval from the Australian National University's (ANU) Animal Experimentation Ethics Committee (AEEC) (Ethics ID: A2017/41; Rodent models and treatments for retinal degenerations). Adult male and female C57BL/6J wild-type (WT) mice (aged between 50-90 postnatal days) were bred and reared under 12 h light/dark cycle conditions (5 lux) with free access to food and water. The C57BL/6J colony was genotyped for the presence of both the Rpe65 450Met polymorphism or the deleterious Crb1 rd8 mutation using previously published primer sets 275,276 . Sequencing for these was conducted at the ACRF Biomolecular Resource Facility, ANU. All animals used possessed the Rpe65 450Met polymorphism but were free of the Crb1 rd8 mutation. Littermate age-matched WT mice were randomly assigned to photo-oxidative damage (PD) and dim-reared control (DR) groups. Animals in the photo-oxidative damage group were continuously exposed to 100k lux white LED light for a period of 1, 3, or 5 days as described previously 265 , with the majority of experiments conducted for 5 days. Dim-reared control mice were maintained in 12h light (5 lux)/dark cycle conditions.

Retinal s-mEV isolation
Mice were euthanized with CO2 following experimental runs. Either two (from one mouse) or four (from two mice -used for high-throughput sequencing) retinas were pooled and collected in Hanks Buffered Saline Solution (HBSS, Gibco; Thermo Fisher Scientific, MA, USA). Retinas were transferred to 500µL digestion solution ((HBSS containing 2.5mg/mL papain (Worthington Biochemical, NJ, USA), 200U DNase I (Roche Diagnostics, NSW, AUS), 5µg/mL catalase (Sigma-Aldrich, MO, USA), 10µg/mL gentamycin (Sigma-Aldrich, MO, USA) and 5µg/mL superoxide dismutase (Worthington Biochemical, NJ, USA)) and finely chopped using scissors. Retinas were incubated at 37°C for 8 minutes, followed by 20 minutes at 8°C, to allow for the breakdown of the extracellular matrix and s-mEV release.
Following digestion, tissue suspensions were neutralized by diluting in 11.5mL of HBSS and centrifuged at 1000xg for 10 minutes at 4°C to remove cells and cell debris. The supernatant was transferred to 14x89mm Beckman Ultra-Clear ultracentrifuge tubes (Beckman Coulter, CA USA) and centrifuged at 10,000xg for 30 minutes at 4°C in a Beckman Coulter Optima XE-100 (fitted with a SW41Ti Rotor (Beckman Coulter, CA USA)), to collect large EVs and remaining cell debris. The s-mEV-containing supernatant was transferred to new ultracentrifuge tubes and centrifuged for 1.5 hours at 150,000xg at 4°C. The supernatant was carefully decanted, and the s-mEV pellet resuspended via tituration for 1 minute in 500µL Ultrapure Endotoxin-free 0.1M PBS (Thermo Fisher Scientific, MA, USA) and used immediately for quantification.
For RNA isolation, the s-mEV pellet was resuspended immediately in 100µl RNAse A (10µg/ml in Ultrapure Endotoxin-free 0.1M PBS) and incubated for 30 minutes at 37°C to digest any RNA contamination. Following RNase treatment, s-mEV RNA was extracted using the mirVana miRNA Isolation Kit (Thermo Fisher Scientific, MA, USA) as per section 2.8.

NanoSight
The size and concentration of s-mEV were measured using nanoparticle tracking analysis on a NanoSight NS300 (Malvern Instruments, Malvern, UK). s-mEV samples were diluted 1:20 (retinal s-mEV) or 1:40 (cell culture s-mEV) in 1ml Ultrapure Endotoxin-free 0.1M PBS to achieve a particle per frame value between 20 and 100. Samples were analyzed under constant flow provided by a syringe pump set at a speed of 35 (equating to ~3.1µl/min; 433 . A total of nine 30s long videos were captured (camera setting: 14) for each sample. The detection threshold was set between 4-5 and was not altered between measurements of the same experiment. The concentration values, modal and mean sizes were exported to Prism V7 (GraphPad Software, CA, USA) for statistical analysis and plotting.

Zetasizer
Dynamic light scattering measurements were performed using a Zetasizer Nano ZS 90 (Malvern Instruments, Malvern, UK). A 500µL undiluted retinal s-mEV suspension in Ultrapure Endotoxin-free 0.1M PBS was prepared, loaded in a low-volume disposable sizing cuvette (ZEN0112, Malvern Instruments, Malvern, UK) and agitated before measurements.
Measurement parameters were set as follows: Material Refractive Index -1.46, Dispersant Refractive Index -1.330, Viscosity (cP) -0.888, Temperature (°C) -25, and Measurement Duration (s) -60. The acquired intensity data was transformed using the General-Purpose Model within the Zetasizer analysis software to generate the size distribution of the s-mEV.

Transmission Electron Microscopy (TEM)
A 30µl retinal s-mEV suspension was placed on a 200-mesh carbon-coated copper grid (Sigma-Aldrich, MO USA) and pre-treated with glow discharge using an Emtech K100X system (Quorum Technologies, Sussex, UK). After 20 minutes, s-mEV were contrasted with 2% uranyl acetate solution for 1 minute, followed by 3 washes in 0.22µm filtered PBS (Thermo Fisher Scientific, MA, USA). Excess PBS was removed by placing a piece of absorbent paper at the edge of the grid. The grids were imaged on a Hitachi 7100 FA transmission electron microscope (Hitachi, Tokyo, Japan) at 100kV. The images were captured with a side mounted Gatan Orius CCD camera (Gatan, CA, USA) at 4008x2672 pixels resolution using a 2 seconds exposure operated through Gatan Microscopy Suite (Gatan, CA, USA). A total of 20 images were captured at 100,000x magnification from four different grids, each containing s-mEV isolated from a different retinal s-mEV preparation (2 mouse retinas/preparation). The images were imported into ImageJ V2.0 software (National Institutes of Health, Bethesda, MD, USA), scale-calibrated and the diameter of approximately 230 s-mEV was measured. The size distribution was plotted in a histogram with 20nm wide bins using Prism V7.0 (GraphPad Software, CA, USA).

Exosome inhibition
Exosome inhibition was performed using GW4869 (Sigma-Aldrich, MO, USA), a known inhibitor of exosome biogenesis and release 434 . GW4869 was reconstituted in dimethyl sulfoxide (DMSO; Sigma-Aldrich, MO, USA) to a concentration of 5mM and used as a stock solution for further dilution in Ultrapure Endotoxin-free 0.1M PBS. Mice were injected with 1.25mg/kg GW4869 via intraperitoneal (I.P.) injection daily for 5 days. 10.3% DMSO in Ultrapure Endotoxin-free 0.1M PBS (corresponding to the final volume of DMSO in GW4869 preparations) was used as a negative control. All mice were monitored daily for signs of distress or sickness.

Retinal Function via Electroretinography (ERG)
To assess retinal function full-field scotopic ERG was performed as previously described 17 . Briefly, mice were dark-adapted overnight before being anaesthetized with an intraperitoneal injection of Ketamine (100 mg/kg; Troy Laboratories, NSW, Australia) and Xylazil (10 mg/kg; Troy Laboratories, NSW, Australia). Both pupils were dilated with one drop each of 2.5% w/v Phenylephrine hydrochloride and 1% w/v Tropicamide (Bausch and Lomb, NY, USA).
Anaesthetized and pupil dilated mice were placed on the thermally regulated stage of the Celeris ERG system (Diagnosys LLC, MA, USA). The Celeris ERG system has combined Ag/AgCl electrode-stimulator eye probes which measure the response from both eyes simultaneously, and uses 32-bit ultra-low noise amplifiers fitted with impedance testing. Eye probes were cleaned with 70% ethanol and then a 0.3% Hypromellose eye drop solution (GenTeal; Novartis, NSW, AUS) was applied to both probes. The probes were then placed covering and just touching the surface of each eye. A single-or twin-flash paradigm was used to elicit a mixed response from rods and cones. Flash stimuli for mixed responses were provided using 6500K white flash luminance range over stimulus intensities from -0.01 -40 log cd s m −2 . Responses were recorded and analyzed using Espion V6 Software (Diagnosys LLC, MA, USA). Statistics were performed in Prism V7.0 using a two-way analysis of variance (ANOVA) to test for differences in a-wave and b-wave responses. Data was expressed as the mean wave amplitude ± SEM (µV).

Optical Coherence Tomography (OCT)
Cross-sectional images of live mouse retinas were taken at 1mm increments from the optic nerve using a Spectralis HRA+OCT device (Heidelberg Engineering, Heidelberg, Germany) as previously described 265 . Eye gel (GenTeal; Novartis, NSW, AUS) was administered to both eyes for recovery.
Using OCT cross-sectional retinal images, and ImageJ V2.0 software (National Institutes of Health, Bethesda, MD, USA), the thickness of the outer nuclear layer (ONL), was calculated as the ratio to the whole retinal thickness (outer limiting membrane to the inner limiting membrane).

Retinal tissue collection and preparation
Animals were euthanized with CO2 following functional ERG analysis. The superior surface of the left eye was marked and enucleated, then immersed in 4% paraformaldehyde for 3 hours. Eyes were then cryopreserved in 15% sucrose solution overnight, embedded in OCT medium (Tissue Tek, Sakura, Japan) and cryosectioned at 12µm in a parasagittal plane (superior to inferior) using a CM 1850 Cryostat (Leica Biosystems, Germany). To ensure accurate comparisons were made for histological analysis, only sections containing the optic nerve head were used for analysis. The retina from the right eye was excised through a corneal incision and placed into RNAlater solution (Thermo Fisher Scientific, MA, USA) at 4°C overnight and then stored at -80°C until further use.

Immunolabelling
Immunohistochemical analysis of retinal cryosections was performed as previously described 53 . Fluorescence was visualized and images taken using a laser-scanning A1 + confocal microscope at 20x magnification (Nikon, Tokyo, Japan). Images panels were analyzed using ImageJ V2.0 software and assembled using Photoshop CS6 software (Adobe Systems, CA, USA).

IBA-1 Immunohistochemistry
Immunolabeling for IBA-1 (1:500, 019-19741, Wako, Osaka, Japan) and quantification was performed as previously described 53 . The number of IBA-1 + cells (a marker of retinal microglia and macrophages) was counted across the superior and inferior retina using two retinal sections per mouse and then averaged. Retinal cryosections were stained with the DNAspecific dye bisbenzimide (1:10000, Sigma-Aldrich, MO, USA) to visualize the cellular layers.

TUNEL assay
Terminal deoxynucleotidyl transferase (Tdt) dUTP nick end labelling (TUNEL), was used as a measure of photoreceptor cell death. TUNEL in situ labelling was performed on retinal cryosections using a Tdt enzyme (Cat# 3333566001, Sigma-Aldrich, MO, USA) and biotinylated deoxyuridine triphosphate (dUTP) (Cat# 11093070910, Sigma-Aldrich, MO, USA) as previously described 286 . Images of TUNEL staining were captured with the A1 + Nikon confocal microscope at 20x magnification. The total number of TUNEL + cells were counted including both the superior and inferior retina using two retinal sections per animal, and is represented as the average number of TUNEL + cells per retinal section.
To further quantify photoreceptor survival, the thickness of the ONL on retinal cryosections was determined by counting the number of nuclei rows (photoreceptor cell bodies) in the area of retinal lesion development (1mm superior to the optic nerve head). Photoreceptor cell row quantification was performed five times per retina using two retinal cryosections at comparable locations per mouse. The thickness the ONL, inner nuclear layer (INL), and the combined ganglion cell layer (GCL)-outer plexiform layer (OPL) thickness were also measured at the lesion site on the superior retina, and expressed as a ratio to whole retinal thickness.

In situ hybridization
Localization of miR-124-3p within the retina was determined by in situ hybridization. A double DIG-labelled miR-124-3p miRCURY LNA miRNA Detection Probe (Exiqon, Vedbaek, Denmark) was used on retinal cryosections, which were hybridized for 1 hour at 53°C as previously described 245 . The bound probe was visualized using 5-bromo-4-chloro-3 indoyl phosphate (NBT/BCIP; Sigma-Aldrich Corp., St. Louis, MO, USA). Bright field images were captured on the A1 + Nikon confocal microscope fitted with a DS-Ri1-U3 color camera at 20x magnification and 4076x3116 pixel resolution. All images were centered at the site of lesion located approximately 1mm superiorly to the optic nerve head. The images were imported into ImageJ V2.0 software, converted to 8-bit format and then the densitometry was

cDNA synthesis from mRNA and miRNA templates
Following purification of RNA, cDNA was synthesized from 1µg RNA using either the Tetro cDNA Synthesis Kit (Bioline Reagents, London, UK) from an mRNA template, or using the TaqMan MicroRNA RT kit (Thermo Fisher Scientific) from a miRNA template, according to manufacturers' instructions.

Quantitative real-time polymerase chain reaction:
The expression of ESCRT-independent exosome biogenesis pathways genes was measured by qRT-PCR. We targeted Pdcd6ip (also known as Alix), which encodes an accessory protein in the ESCRT-dependent pathway, and Smpd3, which encodes nSMase2 in the ESCRTindependent pathway 207 . The expression of miR-124-3p was also investigated in retinal lysates from exosome-inhibited mice, and controls. The expression of these genes and miRNA was measured using mouse specific TaqMan hydrolysis probes ( MA, USA). Reactions were performed in technical duplicates in a 384-well format using a QuantStudio 12 K Flex RT-PCR machine (Thermo Fisher Scientific, MA, USA). Data was analyzed using the comparative Ct method (ΔΔCt) and results are presented as percent change relative to control. Expression was normalized to reference gene glyceraldehyde-3-phosphate dehydrogenase (Gapdh) for mRNA, and small nuclear RNA U6 for miRNA. million reads/sample with an average phred read quality of 33 ( Figure S3 A). Sequencing libraries prepared from whole retinal tissue were retrieved from BioProject database (PRJNA606092). These libraries were previously prepared by our group using the same library construction method, and the same bioinformatic analysis pipeline was applied as stated below, see Section 2.9.2.

Bioinformatics
Sequencing reads were initially checked for quality scores, adapter/index content and Kmer content using FastQC v0.11.8 ( Sequencing data can be accessed from BioProject (Accession ID: PRJNA615966).

Network and pathway enrichment analysis
The miRNet platform 302 was used to elucidate potential interactions between exosomal miRNA and retinal mRNA. The mRNA dataset is available from BioProject (accession ID: PRJNA606092) and comprises of all retinal genes with a log2(count per million) value > -2.4 (Table S2). The retinal targetome of the top 10 exosomal miRNAs as well as the targetome of the s-mEVs enriched miRNA were separately imported into Enrichr 303 and analyzed for overexpressed pathways annotated in Wikipathways 304 (mouse annotation), and gene-disease associations listed in DisGeNet database 305 .

661W cell culture
Murine photoreceptor-derived 661W cells ( Cells were maintained and all incubation steps were performed in dark conditions in a humidified atmosphere of 5% CO2 at 37°C, unless otherwise stated. Cells were passaged by trypsinization every 3 -4 days.
To deplete FBS of s-mEV, the serum was centrifuged (200, Control cells were completely wrapped in aluminum foil with six small incisions to allow air/gas exchange.

Cell viability by MTT assay
Cell viability was tested by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide Background absorbance (wells containing medium with MTT reagent only) was subtracted from all sample wells and the viability calculated by dividing the absorbance of treated wells (DMSO or GW4869) by controls (cell treated with medium only).

Exosome inhibition in 661W cells
To test the effect of GW4869 on exosome inhibition, 661W cells were seeded at a density of 1x10 5 cells/well in 6 well plates (Nunc, Thermo Fisher Scientific, MA, USA) and grown to 80% confluence in growth medium. Media was removed and cells treated with 20µM GW4869, or equivalent DMSO, in adjusted growth media for 4h. The conditioned medium from four GW4869-treated or control wells (661W cells treated with equivalent DMSO concentration) was pooled and processed for s-mEV isolation and characterization as described in Sections 2.2 and 2.3.1.

Statistical analyses:
All graphing and statistical analyses were performed using Prism V7.0, unless otherwise specified. An unpaired Student's t-test, one-way analysis of variance (ANOVA), or two-way ANOVA with Tukey's multiple comparison post-hoc test was utilized as appropriate to determine the statistical outcome. Non-adjusted p values (P<0.05) or false-discovery adjusted p values (FDR<0.1) were deemed statistically significant. All data was expressed as the mean ± SEM.

Retinal s-mEV isolation
In this work we demonstrate a novel protocol for the isolation of s-mEV from mouse retinas ( Figure    PD groups (P>0.05, n=5-6). All results obtained using nanotracking analysis (NanoSight NS300).

GW4869 inhibits s-mEV bioavailability in vitro.
Given the correlation between increased photoreceptor cell death and reduced retinal s-mEV concentration during photo-oxidative damage, the effect of s-mEV, specifically exosome inhibition was investigated using GW4869. To determine efficiency, GW4869 was added in culture to 661W photoreceptor-like cells ( Figure   Scale bar = 50nm. * Patterned bars in Figure 3E represent equivalent dosages/concentrations and conditions of GW4869/DMSO used in Figure 3D.

Endocytic pathway inhibition results in reduced s-mEV bioavailability in vivo.
The contribution of exosomes to retinal s-mEV population was investigated in vivo Additionally, the expression of genes associated with ESCRT-dependent and ESCRTindependent exosome biogenesis pathways were measured by qRT-PCR. We targeted Pdcd6ip (also known as Alix), which encodes an accessory protein in the ESCRT-dependent pathway, and Smpd3, which encodes nSMase2 in the ESCRT-independent pathway 207 . Retinal lysates from dim-reared and 5-day photo-oxidative damaged retinas as well as lysates from GW4869 and DMSO-injected photo-oxidative damaged retinas were used. The expression of Pdcd6ip and Smpd3 was significantly increased in 5 day photo-oxidative damaged retinas compared to dim-reared controls ( Figure  These results demonstrate that GW4869 can be used as an inhibitor of exosome production, and that exosomes likely contribute to the population as well as effects of retinal s-mEV. Nanotracking analysis (NanoSight NS300) quantification revealed a significant reduction in the concentration of s-mEV isolated from both DR and 5D PD retinas following 5 daily intraperitoneal injections of GW4869 compared to respective DMSO controls (P<0.05, n=4).
The expression of exosome biogenesis genes Pdcd6ip and Smpd3 as measured by qRT-PCR was (E) significantly increased following 5D PD relative to DR controls (P<0.05, n=5) and (F) significantly reduced following daily injections of GW4869 while undergoing 5D PD, relative to DMSO injected controls (P<0.05, n=4).

Exosome inhibition reduces retinal function in dim-reared mice
The effect of GW4869-mediated exosome inhibition on retinal health was investigated in dim-reared mice ( Figure  While small extracellular vesicle fractions isolated following high speed >100,000xg ultracentrifugation and expressing tetraspanin markers CD9, CD63 and CD81 (such as those isolated in this work) are commonly referred to as exosomes, without evidence of endosomal origin and in complying with MISEV 2018 guidelines 238 , are herein referred to as small-tomedium EV, or s-mEV. In reference to other works, EV terminology will be referred to as published in the original papers.  (Table   S1).
A total of 154 miRNAs were detected in retinal s-mEV (Table S1)   accounted for approximately 67% of the total counts.

s-mEV miRnome is associated with inflammatory, cell death and motility pathways.
Considering that s-mEV miRNAs were not altered following 5 days of photo-oxidative damage and that the top 10 most abundant miRNAs accounted for 67% of the total s-mEV miRNAome, we focused on these top 10 s-mEV-miRNAs. A network analysis was performed to understand the interactions between the top 10 s-mEV-miRNAs and the retinal transcriptome  Table S4). The predicted targets of both sets of miRNA were separately used for enrichment analyses against WikiPathways (mouse pathway annotation) and DisGeNET (database containing gene-disease associations) on the Enrichr platform. A total of 50 pathways were significantly over-represented in WikiPathways (Table   S5 and S6). Notably, pathways pertaining to inflammatory processes including IL-1 to IL-6 signaling, Toll-like receptor signaling and chemokine signaling showed a significant enrichment. Pathways related to cell survival and motility were also significantly enriched

Discussion
This study describes for the first time, the isolation and characterization of mouse retinal s-mEV and the important role they play in retinal health and degeneration. We demonstrate four key findings from this study: Firstly, we demonstrate that s-mEV isolated from mouse retinas decrease in concentration progressively as a consequence of retinal degeneration.
Secondly, we show that partial s-mEV depletion, via systemic administration of the exosome inhibitor GW4869, resulted in reduced retinal function in the normal retina and further exacerbated functional losses in mice subjected to photo-oxidative damage. Notably, significant photoreceptor cell death and inflammation were observed in GW4869-injected mice undergoing photo-oxidative damage, which was mirrored in vitro by 661W photoreceptor-like cells displaying increased susceptibility to photo-oxidative damage following exosome inhibition. Thirdly, we used small RNA sequencing and bioinformatic analyses to report on the potential regulatory roles of s-mEV miRNAs in retinal degeneration. Lastly, using the s-mEVabundant miRNA miR-124-3p as a measure, we provide evidence that retinal s-mEV could be involved in the dynamic cell-to-cell transfer of miRNA in the degenerating retina. Taken together, we propose that retinal s-mEV and their miRNA cargo play an essential role in maintaining retinal homeostasis through immune-modulation, and that this effect is potentially mediated by s-mEV populations including exosomes.

Photoreceptor cell death is associated with reduced s-mEV bioavailability
Previous studies have demonstrated that miRNA-laden exosomes are found in abundance An additional hypothesis is that in response to excessive light, increased numbers of retinal s-mEV are mobilized as early responders to stress. However, even by 1 day of photo-

Retinal s-mEV as mediators of immuno-modulation.
Regardless of origin, to date the role that s-mEV play in retinal health and disease is still largely unclear. Results from this work strongly support a mechanism by which retinal s-mEV and in particular exosomes mediate homeostasis and immuno-modulation, with the inhibition of exosomes using GW4869 both in vitro, and in vivo, resulting in increased cell death, as well as recruitment and activation of microglia/macrophages. Importantly these observations were only evident under stress conditions, with both control 661W cells and dim-reared retinas displaying no major signs of cell death or inflammation following exosome-inhibition in the absence of photo-oxidative damage. We suggest that unlike in the degenerating retina, that exosome inhibition had no major effects on cell health. This is likely to be a consequence of the experimental paradigm used in this study and the short period of inhibition, or alternatively could indicate that under stress exosomal communication is necessary for cell survival. A smaller average size was seen in 661W-isolated s-mEV compared to those from the retina, and while we attribute this to the heterogenous nature of whole tissue, it could suggest that photoreceptors primarily secrete a smaller s-mEV fraction such as exosomes, and it is this population that may mediate retinal damage. This hypothesis however requires further investigation.
We hypothesize that as a consequence of longer-term exosome inhibition, inadequate translocation of miRNA cargo via s-mEVs results in the dysregulation of immune pathways.
We have previously reported that in response to photo-oxidative damage, miR-124-3p upregulation in the INL may occur via outer-to-inner retinal translocation, with miR-124-3p acting as an anti-inflammatory regulator of C-C Motif Chemokine Ligand 2 (Ccl2) to prevent the recruitment of microglia/macrophages 245 . In this present study, we provide further evidence of s-mEV-mediated miRNA translocation, demonstrating that in mice treated with GW4869, INL upregulation of miR-124-3p, the most highly expressed miRNA in isolated retinal s-mEV, was reduced. Although correlative, we suggest that insufficient gene regulation due to reduced s-mEV/miRNA bioavailability could contribute to the increased presence of immune cells and inflammation as seen in retinal degenerations. While we do not exclude the possibility that miR-124-3p could be upregulated in the INL in response to photo-oxidative damage, and downregulated following treatment with GW4869, we believe that this is unlikely, given the lack of differential change in the expression of miR-124-3p in photo-oxidative  Under normal homeostatic conditions, s-mEV laden with miRNA, including miR-124, let-7, miR-125 and miR-183, potentially originating from photoreceptors, are continuously trafficked between retinal neurons and glial cells. However, in retinal degenerations, a loss of photoreceptors leads to a decrease in s-mEV bioavailability, and consequently an insufficient transfer of s-mEV miRNA cargo to their cell targets. As the retinal targetome of s-mEV-miRNAs is associated with inflammation, survival and cell motility pathways, inadequate pathway regulation following s-mEV depletion leads to exacerbated retinal damage.

s-mEV as mediators of phototransduction
In addition to a potential role as regulators of retinal homeostasis through immune modulation, we highlight a possible involvement of retinal s-mEV, and in particular exosomes, In addition, exosomes derived from microglial cells and injected into the vitreous of mice subjected to oxygen-induced retinopathy showed protective effects, reducing avascular regions in the retina, VEGF expression, and photoreceptor apoptosis, compared to controls 465 . It was hypothesized by these authors that exosomal-miR-24-3p mediated this protection against hypoxia-induced cell death 465 .
The combined findings of our work uncover a novel role for retinal s-mEV in both health and degeneration, unveiling a panel of s-mEV-miRNA required for retinal homeostasis, and target networks of these gene regulators comprising inflammatory, oxidative stress and cell survival pathways. As we elude to retinal health requiring optimal levels of s-mEV, and their cargo; replenishing s-mEV loads in the retina itself may prove as efficacious therapy, and will be the focus of future works. Further, both the unique s-mEV-miRNA signature and downstream target pathways open additional avenues for therapeutic development.

Conclusion
Results from this work suggest that s-mEV are released from photoreceptor cells to maintain retinal homeostasis. However, as a consequence of photoreceptor cell death, s-mEV secretion and/or bioavailability becomes reduced. Consequently, retinal s-mEV cargo, which contains regulatory miRNA and other molecules, are unable to regulate immune responses, subsequently contributing to progressive retinal cell death. We hypothesis that this mechanism is likely to be involved in many retinal degenerative and inflammatory diseases.